A hierarchical framework for improving ride comfort of autonomous vehicles via deep reinforcement learning with external knowledge

Ride comfort is an important consideration in the public acceptance of autonomous vehicles (AVs). Road profile, driving speed, and the suspension system can affect AV ride comfort. The authors propose a hierarchical framework to enhance ride comfort by integrating speed and suspension control in vehicle-to-vehicle and vehicle-to-infrastructure communications. Based on efficient speed planning through dynamic programming, the authors propose a suspension control, based on machine learning, that adapts to changing pavement conditions. A deep deterministic policy gradient with an external knowledge (EK-DDPG) algorithm is designed for efficient self-adaptation of suspension control. External data of action selection and value estimation derived from other AVs are combined into the loss functions of the DDPG algorithm. To verify the method, it is applied to actual pavements detected in districts of Shanghai. The results show that the EK-DDPG-based suspension control improves ride comfort on rough pavements by 27.95% and 3.32%, compared to a model predictive control (MPC) baseline and a DDPG baseline, respectively. Also, the EK-DDPG-based suspension control improves computational efficiency by 22.97%, compared to the MPC baseline, and performs at the same level as the DDPD. The authors’ study provides an efficient method for improving the ride comfort of AVs.

• Record URL:
  • http://dx.doi.org/10.1111/mice.12934
• Availability:
  • Find a library where document is available. Order URL: http://worldcat.org/issn/10939687
• Authors:
  • Du, Yuchuan
  • Chen, Jing
  • Zhao, Cong
  • Liao, Feixiong
  • Zhu, Meixin
• Publication Date: 2023-5-15

Language

• English

Media Info

• Media Type: Web
• Features: Figures; Maps; References; Tables;
• Pagination: pp 1059-1078
• Serial:
  • Computer-Aided Civil and Infrastructure Engineering
  • Volume: 38
  • Issue Number: 8
  • Publisher: Blackwell Publishing
  • ISSN: 1093-9687
  • Serial URL: http://onlinelibrary.wiley.com/journal/10.1111/(ISSN)1467-8667

Subject/Index Terms

• TRT Terms: Active suspension systems; Autonomous vehicles; Machine learning; Ride quality; Vehicle to infrastructure communications; Vehicle to vehicle communications
• Subject Areas: Data and Information Technology; Highways; Safety and Human Factors; Vehicles and Equipment;

Filing Info

• Accession Number: 01891653
• Record Type: Publication
• Files: TRIS
• Created Date: Aug 28 2023 9:34AM